feat: broken link checker script

[bzp2010]

Fixes: #255

Changes:

A broken link checking script is implemented using NodeJS. It can scan the link and send an HTTP request to check if the link is available.

Problem:

Currently, markdown link detection uses an external library, which will have problems in link identification when checking links without title. Then consider refactoring this part of the code.

[bzp2010]

Updated 2021-3-18

• refactoring link scanner
• create workflow

Tips

• only links starting with http:// or https:// are checked
• relative paths and page anchors are not checked
• the list of broken links is stored by JSON in workflow artifacts

Problem

Is it necessary to check the existence of files linked by relative paths?

If necessary, the current PR has not been completed.

[juzhiyuan]

Ya, I would prefer checking those links referenced across files.

[bzp2010]

OK, I will continue to work on it.

[bzp2010]

Updated 2021-3-19

• add local markdown file checking function

Tips

• links starting with http:// or https:// and ending with .md are checked

Problem

Is there anything else to add?

[juzhiyuan]

I noticed that there have some files end with .lua ?

[bzp2010]

Those .lua files were not copied to the document directory, which prevented it from being checked. Therefore, it will be regarded as a broken link.

[juzhiyuan]

yes, could we redirect users to the original site? (apisix's GitHub I mean)

[bzp2010]

yes, could we redirect users to the original site? (apisix's GitHub I mean)

Redirect no .md files to GitHub page?

[juzhiyuan]

.lua to github original file 🤔

[bzp2010]

.lua to github original file 🤔

I need to try. Due to the problem of not copying other files, the correct file location relationship may not be established.

[bzp2010]

Updated 2021-3-20

• add link conversion and check function for non document files

Tips

• non document files link will be converse to a web link on GitHub

[bzp2010]

Summary

• The target function has been realized. (request review)
• In order to ensure the correctness of the document, the link conversion function is not introduced in the workflow involving the output of compiled products. (Do we need this?)
• The current implementation is not elegant enough. (Need help for improvement)

Thanks

[bzp2010]

The common code in the link-checker module has been splited. (such as document path generation)
The code in the document synchronization module will be implemented in the new PR.

[juzhiyuan]

LGTM